[CF-metadata] Accumulation since timestep 0, cell-method

[Jonathan Gregory]

Dear Heiko

This is an uncommon case, I would say. I think your bounds solution is correct,
but it would be recommended also to include a cell_methods of "time: sum". I
agree this means you will need a separate time-variable for this quantity, but
that may be useful because you might like to give it different coordinate
values as well: perhaps they would be either the end of each accumulation
period or the middle of it, whereas the "ordinary" time-coordinate appropriate
for time-means would probably have values of 1.5, 4.5, 7.5, ...

Best wishes

Jonathan